Netflix or Amazon Prime push you like watching movies, the logic behind this is that you do not feel surprised? Or do not you curious what makes Google Maps can predict the traffic on your travel route?
We all know how to use machine learning algorithms and statistical models to perform the task and proposed the perfect solution. Also, this method can detect cancer, and a variety of purposes and to help detect faces on Facebook.
Machine Learning: Demand
Machine learning algorithms and their daily imitate the law of human development. In short, machine learning can be divided into two concepts: training and forecasting.
Machine learning has appeared in our daily lives, but we are hardly aware of. For example, on social media platforms to the user tag is just a machine learning simple tasks only. Machine learning applications such as widespread fraud detection, identification and recommendation systems. In the near future, machine learning will be used in self-correcting, providing insightful values and personal service these technologies.
How machine learning algorithm works
Machine learning is created that can answer every question raised by users of the system. The system then modeled through training most appropriate algorithm and use it as a basis to answer questions.
Rather, a machine learning model has seven steps to follow:
From the escalator detection requires immediate repair to detect skin diseases, machine learning, gave birth to the computer system that can magically deal with some things we can not understand. But how machine learning work? In the absence of explicit programming to the situation, what steps will be taken and how they work? This is what you need to know.
Here, we will demonstrate the working principle of machine learning by reference to an example: We get beer and wine, for example, through which you can create a system, the system will respond to a given beverage is wine or beer.
1, data collection
Here you can give a simple example. Data to be collected is obtained from a glass with a beer or wine in. From the shape analysis to check the quantity of glass foam, data collection can be anything. Here, the color of these liquids selected wavelength of light, and the contents (alcohol) as a feature. The first and most important step, including the purchase of several types of alcohol from retail stores, as well as the device can be equipped with the correct measurements, for example for measuring the color spectrometer, as well as for measuring the alcohol content of the hydrometer.
This step is important, because the quality and quantity of data collected will help to improve the accuracy of the prediction model. Alcohol content and color of each collection of beverage alcohol or wine is to find ingredients, and we are ready for this system is the same training data system.
2, data preparation
Once collected the data, you need to load it into the system, and to prepare for machine learning and training.
These data are randomly placed, so the system does not know a drink is part of the beginning of beer or wine. But the system should be able to identify the beverage is wine or beer. At the same time, to visualize the operation to ensure that no imbalance exists between the variables.
However, if we collect data beer than wine, then the trained models may show a deviation of beer to some extent, because most of the data collection is all about the beer. However, in real-time situation, if the model using both an equal number of data beer and wine, then beer prediction might be wrong half.
Therefore, to provide the correct amount of data is also important for the two variables.
3, select the appropriate model
How do I know which model is appropriate? According to a number of researchers and data scientists, it is clear that the experts will choose the correct model have their own ideas.
For example, some models have been designed, most suitable for the musical sequence or text or the like, while others suitable sequence of numbers. Beer and wine in our example, it is a linear model, as you will see beer and wine are two different characteristics.
4, the training model
which is a crucial process because it uses the data to further improve the performance of the model - predicted wine and beer.y=m*x+b
y is the intercept, m is the slope of the line, y is the value of the linear position x, b is a straight line in the X-axis intercept. m, b, and y is the only value of training and evaluation.
In machine learning, you will encounter more than m variables from which you can construct a matrix or weight matrix w.
5, evaluation
Next is the evaluation, the evaluation process to check whether the model of effective training or whether the job done. In this way you can easily use the data has not been seen in training to test the model. The idea is to test how the model response data has not yet encountered. Ideally, the evaluation is to analyze how the model performed in real time.
6, ultra-parameter adjustment
This is to check whether the room is being trained models still room for improvement. It can be achieved by adjusting certain parameters (learning rate, or the number of times in the training process training model run).
During training, you have to consider multiple parameters. For each parameter, you have to know them in the model training role, or you may find yourself wasting time or after parameter adjustment through time-consuming longer.
7, forecast
The final step, once followed the above parameters, the model can be tested. Given color and alcohol content, the machine can predict what kind of drink beer and wine. Machine learning can be used instead of the standard model by means of rules or human judgment to determine a difference between wine and beer.
Known machine learning applications
Even before we realized that we had been using machine learning, this is incredible. As we all know, machine learning has applications in a variety of industries, such as medical diagnosis, speech recognition, learning associations, financial services, prediction.
Medical diagnosis
Machine Learning provides the tools and technologies can benefit the medical field, it helps to solve the disease predict and diagnose problems.
It has also been used to analyze clinical parameters used for disease prediction, for example, it helps predict disease progression, but also help to improve the treatment plan, mainly for patient management in general.
Speech Recognition
In speech recognition, machine learning will help convert spoken words into text that automated speech recognition or voice into text or computer speech recognition.
Associative learning
This is an insight to the development of the association between the product of the process. In short, unrelated products can also reveal associations between them.
Financial Services
Machine learning system is a good tool, through continuous monitoring of individual activity to detect fraud and to assess whether the activities of individuals belonging to this user.
prediction
Machine learning can predict the likelihood of the customer loan defaults. However, for computing, the system to be classified for a particular set of data.
Original link: https://imba.deephub.ai/p/18f788a063fe11ea83daff02656c39f6
Scanning micro-channel two-dimensional code, more dry